ECS Flashcards
How can containers access ports on the host container?
Port mappings are specified as part of the container definition, which can be configured in the task definition
What information is in the ECS task definition?
- Image Name
- Port Binding for Container and Host
- Memory and CPU required
- Environment variables
- Networking information
- IAM Role
- Logging configuration (ex CloudWatch)
How many containers can you define per task definition?
10
Explain load balancing with EC2 launch type
- you get a dynamic host(instance) port mapping if you only define the container(task) port in the task definition
- the ALB is able to find the right port on your EC2 instances
- you must set up the Security Group on the EC2 instance to allow any port from the ALB security group (i.e., has to allow the ALB to connect via any port)
How many roles per task definition should we have?
one
What is the hierarchy of stuff in ECS?
Service
└── Task Definition (specific version)
└── Container Definition(s)
└── Running Containers (in Tasks)
What are some common patterns for ECS task definitions with multiple containers?
- Main Application + Sidecar = web app container + logger container sidecar
- Web Application + Redis Cache: web app needs local caching so both containers need to be on the same host
- Application + monitoring agent need to be on same host
- API + database proxy: db connection pooling needs to be close to the app
What are the key reasons to use multiple containers in a task?
Containers need to share network namespace
Containers need to share storage volumes
Containers need to be scheduled together
Containers have a strong lifecycle dependency
Containers need low-latency communication
When should you not use multiple containers?
Independent services that can scale separately
Services that don’t need to share resources
Services that can tolerate network latency
With ECS, if you prefer be charged per running task rather than running container instances, which model should you use?
Fargate
